EXPLORING THE BEHAVIOR OF NEURAL-NETWORK SOFTWARE QUALITY MODELS

In the paper, we explore the behaviour of neural network software qual ity models, We select a multiple regression quality model from the pri ncipal components of software complexity metrics collected from a larg e commercial software system, We train two neural networks; one with t he complete set of principal components and one with the set of compon ents selected by multiple regression model selection. Comparisons of t he three models, for two quality measures gathered from five related s oftware systems, lead to a better understanding of neural network soft ware quality models, and the relationship between software complexity metrics and software quality metrics,